I’m feeling a little better about the threat to my writing business posed by LLMs:
Late last week, MSN.com’s Microsoft Travel section posted an AI-generated article about the “cannot miss” attractions of Ottawa that includes the Ottawa Food Bank, a real charitable organization that feeds struggling families. In its recommendation text, Microsoft’s AI model wrote, “Consider going into it on an empty stomach.”
…the article extols the virtues of the Canadian city and recommends attending the Winterlude festival (which only takes place in February), visiting an Ottawa Senators game, and skating in “The World’s Largest Naturallyfrozen Ice Rink” (sic). [Links in the original]
More seriously, I read an interesting analysis suggesting that the AI craze, at least as an economic proposition, is very possibly as much bubble as boom. Writing in his Substack, Gary Marcus sees a lot more techno optimism (or outright huckster-hype) than an actual path to the revolutionary impact AI boosters have told us is just about within our grasp:
Everybody in industry would probably like you to believe that AGI is imminent. It stokes their narrative of inevitability, and it drives their stock prices and startup valuations. Dario Amodei, CEO of Anthropic, recently projected that we will have AGI in 2-3 years. Demis Hassabis, CEO of Google DeepMind has also made projections of near-term AGI.
I seriously doubt it. We have not one, but many, serious, unsolved problems at the core of generative AI — ranging from their tendency to confabulate (hallucinate) false information, to their inability to reliably interface with external tools like Wolfram Alpha, to the instability from month to month (which makes them poor candidates for engineering use in larger systems). [Links in the original]
My biases make me hope so, which means I don’t trust my assent to Marcus’s argument. But even (trying to) allow for that thumb on the scale, it doesn’t seem implausible that the impact of LLMs and related AI approaches might be much more an incremental advance in the automation of various tasks (maybe even a big increment) and not the everything-is-different kind of transformations like the chemical-energy and (original) computational ones.
Over to you. I’m off to grab my Michelin starred grocery bag from the food bank.
This thread is as open as Boston Common.
Image: Jan Steen, The Fat Kitchen, between c. 1665 and c. 1670
Frankensteinbeck
99 times out of a hundred, any product that is being hyped for what it will eventually be able to do rather than what it can do now is a fraud.
EDIT – Second!
Ishiyama
Am I too old if I can remember the New Yorker cartoon with the caption: “On the internet nobody knows that you are a dog”?
TeezySkeezy
@Frankensteinbeck:
You look first to my lyin’ eyes.
Sister Golden Bear
AI has good potential within limited domain areas, e.g. helping programmers on a SaaS platform find help articles related to problems they’re having, or recommending training classes. The content a LLM ingests is curated (i.e. it digest only that company’s knowledge base articles training content, etc), and the scope of questions are more limited.
In the broader world, not as much….
misterpuff
I assume most of the customers of the Food Bank are coming in with an empty stomach (or at least empty wallets).
Thank Dog for the food banks and their supporters.
Ohio Mom
Admittedly, I am not searching out examples of AI writing but the ones I have come across by happenstance, like this one about the food bank, are cringe worthy. What I’ve seen so far makes a term paper cribbed from Wikipedia look thoughtful in comparison.
MisterDancer
There are aspects of AI, like writing code, that I think will be deeply sticky. There’s so much open source code to train off of, and so little need to be creative, AND so much need to write code en masse, that I can’t imagine AI-written code won’t have a major impact.
But. I don’t want to call this all a costly “flash in the pan,” but…there’s just a lot of text and art trained against stuff with copyrights and licenses. And the output really skirts the edge of Fair Use at best, given the output we tend to see where we tell AI to “be creative”. Not sure what the outcome of all this is, just that this feels…off, as something that across the board’ll be a game-changer in our culture.
craigie
My own experience has been that the more I use it, the less impressed I am. That’s especially true when I ask for code samples, but all of it is about as good as a self-driving taxi – which is to say, not very.
SpaceUnit
As a bot myself I say you dismiss us at your own peril.
Ohio Mom
@Sister Golden Bear: That sounds right to me.
I am reminded of the joke, “In his head, an original thought would die of loneliness.” There are no original thoughts in AI, at least not yet. Just mashups of prior knowledge.
Scout211
@Ishiyama: I didn’t remember that, so I had to search it.
The cartoon has its own page on Wikipedia.
jackmac
I’d like to turn AI writing loose on Washington politics and see if readers could tell the difference between computer generated horse race political coverage / both-siderism and what’s currently practiced by our so-called media elites.
AI gets the edge at this point because the real people can’t get much worse.
MisterDancer
Hilariously, as I was saying, that’s what a lot of the devs I know are asking for more of. I don’t know if I’m up for using it in that way, but I’m more comfortable leveraging against what I expect is mostly open source code to train against.
Yarrow
An enterprising restauranteur may open a new restaurant in Ottawa called Food Bank.
Pika
Ben Collins has been making the AI bubble point for a while now. It helped to steady and remind me about the whole higher-ed MOOC frenzy from a bit back: EVERYBODY said it would be the big and only thing and then…
scav
Driverless Car Gets Stuck in Wet Concrete in San Francisco FTFNYT
misterpuff
If, in any creative profession, there is a hierarchy: Artiste, Pro, Apprentice, Hack, I can’t see AI replacing any but the Hack. Sure you can output many pages of words at a lower cost, but you’ll still need an editor, etc.
But David Brooks must be quaking in his boots…..
Omnes Omnibus
FWIW I see it as something that could, if used correctly, make legal research easier but not as something that can do legal research.
Jacqueline Squid Onassis
The chances of having a real AGI before having a safe, reliable self-driving car are so small as to be on the Planck Scale. It doesn’t look like we’re close to having a safe self-driving vehicle. AI (in any meaningful sense) is infinitely more complex than a self driving vehicle.
Salty Sam .
This has been my thinking for awhile now. We are not gonna see a HAL9000 AI anytime soon.
Open thread? I have a bleg-
For the past 10 weeks I have been dealing with a sick sister-in-law. She is the sister of my first wife (deceased), and although I never really got along with her that well, she was a motherly presence in my two young sons’ lives after their mom passed, and she means a lot to them. Since I (and my wonderfully helpful now-wife) are the only family in town, her care has fallen to us. I’m doing this mostly for my boys.
In February, she began acting strangely, and my boys were concerned. She was making plans to sell her house and move to Costa Rica- all well and good, except her planning seemed… sketchy. Any questions about the details of her plans were rebuffed, and the red flags were popping up everywhere. I am not a psychiatrist, but I also wasn’t born yesterday- I was recognizing classic signs of a manic phase of a bipolar outbreak.
To make an extremely long story as short as possible, this continued (with increasing alarm for everyone) until May, when she had some sort of breakdown, which included her recognizing that she had just wrecked her life- sold her car and sent the proceeds to a sketchy real estate person in Costa Rica, quit her job and burned that bridge, gave away all her clothes and most of her belongings. We got pressed into service to help when she became frozen with fear and despair at how badly she had shit the bed.
She has experienced some sort of psychic break with reality, suffers from some really weird delusions, and is basically helpless. My now-wife, who has the helper genes fully expressed, jumped in, helped her sell her house, found an apartment she could afford, and got her moved in. We found a psychiatrist/social worker who takes Medicare, and have been to her a few times, but we are not impressed. The ONLY tool in her kit is medication, but we haven’t found anything that seems to work. (Psychiatrist diagnosed this as bipolar disorder).
So my sis-in-law sits in her apartment all day with her cat. She refuses to bathe or wash her clothes. I realized she was out of litter for the cat box, and when I brought a new bag over, discovered that she hadn’t cleaned out the cat box in over a week. She has an extremely flat emotional aspect, there is no joy or happiness in her. At the same time, she is overwhelmed with irrational fears about her situation.
My ask is that if any of you fine jackals have had to deal with a similar situation, what resources did you discover that worked? We are looking for a caretaker to make home visits (this situation has totally consumed our lives since June- we HAVE to back out for our own sanity). We have found that UT med school has a “bipolar lab”, and we are working to get her admitted to that. Beyond that, we are at the end of our rope, and could use some advice or suggestions.
Thanking y’all in advance.
ETA- a bit more detail of her condition: it has been my opinion that she’s been bipolar most of her adult life (one of the reasons I haven’t gotten along with her that well), but she self medicated- she has been a heavy daily pot smoker her entire adult life. But in February, when the Costa Rica plans kicked off, she gave up cannabis, became a vegetarian… made a whole bunch of radical changes in her life. And here we are…
zhena gogolia
@Pika: What a blast from the past!
MattF
I’m sorta looking forward to the Singularity. At least, as depicted in the webcomic QuestionableContent, AIs are smart, funny, and sexy— and determined not to make the same dumb mistakes as humans.
HeartlandLiberal
@Ishiyama: another favorite was the dog scientists standing in front of the dog security officer, who was saying “I know who you are, but I still have to sniff your butts.”
Tom Levenson
@Salty Sam .: I’m very sorry about your sister in law. That kind of break is just terrible to witness, much less to have to provide care for. (Twice in my career colleagues had similar breaks, and my much more limited involvement in their situations was tough enough. Your circumstances are much worse.)
I have no real advice beyond what you already know: you can’t navigate this on your own, and the urgent task is to find out who in your area is on top of such crises.
C Stars
There’s a stupid, cute kind of creativity to it. I really like “naturallyfrozen.” Most people will most often recognize whether something is naturally frozen based on the context, but the idea of squishing together into one word a two word phrase is kind of fun. And looks cute. Like my naturallycurly hair and momjeans.
pika
@zhena gogolia: I confess to having you in mind when I wrote that comment. It’s refreshing to find how in that disaster, at least, it now reads as past blast :)
RedDirtGirl
@Salty Sam .: I don’t have anything to offer except my sympathy. Sounds like an extremely challenging situation that you have handled so graciously. I do have faith in the jackaltariat coming through with some ideas though. We contain multitudes.
Holding your family in the light!
Tokyokie
@Ohio Mom: A fellow former copy editor recently posted a link to an AI-generated story about a long-lasting marriage between country superstars Tim McGraw and Religion Hill. But even howlers like that are not going to improve the demand for people with editing skills, which is why I’m now a nurse.
gwangung
Yeah, not surprised that studios were pinning their hopes on AI…they’re about as creative and insightful as the middling AI engine….
Eolirin
Generative AI in its current form is unlikely to be transformative.
But if you’ve been paying attention to the AI field in general there’s no reason to think that the LLMs that are currently all the rage will be the current state of the art in 10 years and maybe not even in five. There have been big jumps in capability as new methods have been discovered, sometimes in iterative fashion building off the previous advances and sometimes through entirely new approaches. The rate of change and advancement has only been accelerating too.
Given the advances in gpu-driven super computer clusters, hardware is less and less of a limiting factor and that’s only going to speed things up even further.
Is there a bubble right now? Yeah, almost certainly. But if that bubble deflating in the near term, which it almost certainly will, doesn’t result in certain key companies pulling back entirely and universities losing the ability to continue to fund research then where we are in a decade will look very different than where we are now. Because this stuff isn’t about a specific technique or piece of technology in the long run, it’s about a constellation of approaches and the hardware support necessary to enable them. And that’s not going to stop just because we run into limitations with one way of doing things.
Captain C
@jackmac:
If you did an AI writing bit for Bobo no one would believe it was him because the piece would be more coherent, fact-based, and empathetic than he’s capable of.
Mallard Filmore
YouTube offered up this to me:
https://youtu.be/EUrOxh_0leE
title: “AI does not exist but it will ruin everything anyway”
The talking head is an astronomer and has an engaging way of speaking.
currawong
Talking of ‘stupid AI tricks, Murdoch has embraced it here by replacing its pseudo-journalists with AI genertaed articles – riddled with errors. I don’t think anyone noticed a drop in quality,
https://www.theguardian.com/media/2023/aug/01/news-corp-ai-chat-gpt-stories
jackmac
@Captain C: Indeed!
Ohio Mom
@Salty Sam .: Have you checked with your local Council on Aging? (I could be wrong but I think every place has one). They are a clearing house for services for seniors — transportation, housing, Meals on Wheels, that sort of thing. They will probably have a lead on home-maker/personal care aides.
Have you contacted your local NAMI chapter? In my “journey” (hate that Hallmark card description) as an autism mom, I have found other families to be my best source for locating services and supports (and also, what support providers to avoid).
Finally, it can take a while to find the best psychotropic drug combo, don’t give up on that yet.
Good luck, you are doing something very important. Keep us posted.
Eolirin
@Ohio Mom: What is an original thought even? Humans aren’t special. We’re also just doing processing on a bunch of data that come in through admittedly more kinds of inputs than computers currently have access to, and regurgitating outputs based on those transformations. We’re not magic. That process is replicable with enough processing power (which we still don’t have yet, but we’re getting there)
VOR
A company I follow actually does serious work in the machine learning space which borders AI. Their CEO talked about their AI/Gen AI initiatives in their recent earnings call. In the Q&A, one of the Wall St. analysts complained that the CEO didn’t talk about AI early enough in his prepared remarks. Not that the substance was lacking, just that it wasn’t the first topic discussed.
BeautifulPlumage
@Salty Sam .:
Only thing I can suggest is trying to find a different professional, which may not be possible where you are. Also sounds like she may need inpatient care to get her stabilized and back to taking care of herself, if Medicare covers anything like that.
So sorry you’re going through this!
Betty
@Salty Sam .: Based on my very limited experience with someone with severe bipolar disorder, it took trying a number of different medications to find one that helped even out the highs and lows. It is such a challenging condition. I wish you luck and hope you can get her into the program you mentioned.
NotMax
GIGO is not a myth.
CaseyL
@Salty Sam .: On a very practical level, get in an application for her for SSDI, since it does not sound like she’s able to be employed. Work with her medical providers to document and confirm her condition. Everything I hear about SSDi is that it’s a long, hard slog, so you might also try to find someone who can help you navigate the system.
Bipolar is a complicated disorder: I’ve known a few people who have it, and the hunt for a medication regime that works is also a long, hard slog. They usually wind up having to take multiple medications, and deal with the side effects of those meds.
My impression, though, is that medication is absolutely necessary: I don’t know of any counseling that is effective at handling the disorder, per se. Mostly (I think) counseling is about learning how to recognize and deal with the mad surges and abyssal depths. (Like, “Know that you shouldn’t make any plans to relocate when you feel this way.”)
You and your wife are saints for taking this on.
Tokyokie
@Salty Sam .: Some mental-health facilities have programs for day patients. You might see whether anyplace nearby has such a program and whether to can take patients. (The slots are usually limited.)
Tom Levenson
@Eolirin: I broadly agree with you: I think developments in the broad area inadequately named AI or machine learning will continue and that surprising, perhaps wholly unanticipated results will follow over time.
But the idea that we have a culture-defining transformation in the interface between humans, computation and the world staring us in the face right now? Not so much.
And I don’t know what the social and cultural implications of a series of incremental changes in the relationship between networks and software on one side and people on the other will be. No idea.
Ked
At my last job I was working with a junior tech who had been running down a networking issue for a couple of days and had rebooted a core switch (breaking everything in sight) at which point I got dragged into the mess. I asked the tech why they had done that, and they literally said they had asked ChatGPT for what they should do next.
This was a kid who I thought was a bright, rising star of their tech team. But somehow “AI sez so” is apparently enough to override all common sense. There were a whole bunch of things wrong with the MSP model as well, but that rant is a little off topic here.
Ultimately, AI is powerful in narrow directed slices now, but it cannot generate meaningful output outside of the zone in which it has been deliberately trained. The basic human experience is the collision of contexts and meta layers, and even the most basic queries we make using casual language will blow through whatever training envelopes are established. AGI is decades and decades away. Anyone who thinks that implementing AI as a replacement for human work is possible is going to pay the price as everything blows up in their face.
RSA
I’m an AI program manager in a government research lab, which means I have something of an insider’s view of the field. A few observations and opinions:
Ohio Mom
@Eolirin: I sometimes think that of science and technology. There are natural laws that someone was going to describe, if Galileo hadn’t discovered the things he did, eventually someone else would have. Someone was going to figure out how to build an airplane, if not the Wright Brothers, someone else would have. Maybe it would have taken longer.
But the arts? No art movement was inevitable. Impressionism wasn’t waiting to be discovered, there wasn’t a violin that was waiting to be developed — who knows what other instruments could be built and give us novel sounds?
*Those* are original thoughts.
And while it’s true that what we are able to see is in large part determined by our culture, look at all the cultures in the world. It’s dizzying.
Salty Sam .
Heh. My wife has shouldered most of the load- my eldest son told her she has earned her sainthood merit badge 5 times over. I concur.
Another Scott
+1
I’m reminded of a seminar I saw during a summer job at WPAFB in the very early 1980s. An Air Force researcher was working on a generalized optical character recognition system. One that would be font-independent. His particular task at that time was to “define B-ness”. IOW, what makes a scratch on a page a letter “B”?
Fast-forward 30-40 years and OCR is pretty good. But not anywhere close to perfect. There’s still lots of human hand-holding required to fix the errors and make a scanned document as good as a human.
I think this LLM AI stuff is at a similar place. It can be good, but it can’t be trusted without verifying it. Nobody would take an OCR of some 600 year old text at face-value without thoroughly checking it. People who are thinking that LLM AI is actually good in its present state, and doesn’t need to be checked, are buying much too much hype.
Progress might be faster than with OCR, or might not. Who knows. But one has to figure that good LLM AI is not going to be cheap (just like pills that cure cancer are not cheap)…
Cheers,
Scott.
Boris Rasputin (the evil twin)
@Frankensteinbeck:99 times out of a hundred, any product that is being hyped for what it will eventually be able to do rather than what it can do now is a fraud.
I call these “Messiah” systems: all-powerful, all-wonderful, the answer to everything. When they arrive, late and over budget, instead of performing miracles, they fumble card tricks. “But it’s going to get better soon!” Bullshit.
waspuppet
My fear isn’t that AI will be good enough to replace writers. My fear is that the people who employ writers won’t care and just use it anyway.
Litlebritdifrnt
I have had one hell of a two weeks with regards to my health and I feel the need to tell you about it because of all the negative press that the NHS gets (particularly the US that loves to demonize “socialist” health care). It is going to be long so please feel free to ignore it if the subject is of no interest.
As you know I was diagnosed with lung cancer about two years ago. It was a small tumor and not aggressive so for the past 2 years I have been monitored by the onchology team with a CT scan, blood tests and consultation with the team every three months. Nothing changed. The last scan showed that the tumor had grown by about 1cm, and so the decision was made to target the tumor with radio therapy (as opposed to whole body chemo or radiation). To that end I was to have a PET scan and a lung function test. In the interim Hubby and I rescued a 3 year old Nano Bully (basically miniature bulldog) called Toad. Anyway at some point after getting her I attempted to pick up up onto my chair and hurt my back, thinking shoulder muscles area cause that is where it hurt. I took some pain killers and it seemed to be getting better. I went for the PET scan and on getting off the scanner table I felt more pain, but me being me ignored it. That was a Wednesday, by the Saturday the pain was increasing and I began going into spasms. I went to the ER and after several tests was diagnosed with a kidney infection and/orUTI (Given that it was “classic” kidney pain). Given anti-biotics and pain killers and sent home. By Sunday I could no longer function, the spasms were so bad I was contorting wildly. Back to the ER. Eventually they reviewed the PET scan and discovered that the cancer from my lung had weakened one of my vertebra (T-10) and it had fractured, fluid was leaking in my spine. I was admitted, whisked to a ward, a catheta was put up my flue and I was immobilized. Numerous scans ensued and of course every time they had to transfer me from one bed to the next the danger was that there would be more damage. On Thursday I was taken for Radio Therapy to zap the vertebra in question and the onchologist had warned me that the pain would get worse before it got better. He wasn’t wrong. On Thursday and Friday if someone had given me a gun I would have quite happily shot myself. The team (Onchology, Cardiology and Ward Doctor plus all their nurses) were trying to figure out the best cocktail of drugs to stop the pain. They were squirting liquid morphene into my mouth, IV drugs, pills, you name it. By Friday afternoon they had the mix just right and the pain stopped. At that point I was on liquid morphene, morphene pills, diazapan, paracetamol, and lidocane patch. By Tuesday I was stable and able to go home once they established that I had a support network to go home to (for me that in addition to my husband four out of five adults in my family are health care professionals). Yesterday I was left alone to settle in then this morning it began. First call was from Onchology, telling me the next steps. As I am now stage 4, Radio Therapy is not going to be effective so we are going to chemo, I see them in a couple of weeks to get that started. Next my local GP called, and made an appointment to come and see me this afternoon (around 4pm). Next call was from Macmillan Cancer support team, they went into action getting all the other teams into place as well as apply for my disability blue badge for the car. GP came round with his nurse and went over a bunch of other stuff with me, made an appointment with the specialist care nurse to come and see me next Wednesday. Checked me out (they are all worried that the spinal fluid leak could cause paralysis so he was checking the strength of my legs etc) bowel movements all that neat stuff. Occupational Health and Community Services are coming round on Saturday to ensure I don’t need any modifications to the house (I don’t an evaluation had been done a couple of years ago when I was ill), they are going to replace my bed with a “hospital style” adjustible bed to help me get up in the morning and fit a modificaton to my toilet so it is raised up and easier to sit down on. They also made appointments for my newly assigned District Nurse to come and see me and physio tharapy to come round as well. In the space of 24 hours an entire army of healthcare professionals are decending on my little bungalow to look after me, little absolutely insignificant (other than to my family) me.. Every single one of them have been brilliant. The nurses on the ward were not just doing their job, they cared for me, they allowed me to almost break their hands in some cases when I was in the worst of my spasms and they didn;t leave my side, not once. Oh and so far my bill has been a whopping £3.05p because I bought a bag of potato chips and a fruit drink from the ward trolly.
This is what healthcare is supposed to be. I knew the NHS was brilliant, and I knew, like many other Brits that the conservatives have been trying to dismantle the entire system in order to sell it off to their rich friends in the insurance industry for profit. It must never be allowed to happen, ever. After this last two weeks I will do everything in my power to ensure that we hold on to our absolute jewel in the crown that is the NHS, if it wasn’t for them I wouldn’t be typing this now. I will never forget it and I will never allow the people in power to forget it either.
Thanks for listening.
Ohio Mom
@CaseyL: If SIL is on Medicare, she is probably old enough for Social Security retirement, or close to it. I am not sure, but do not think you can get disability and retirement at the same time. This is the sort of thing the Council on Aging can help with.
Agree with you that proper medication is key and that Salty Sam and wife are saints.
Omnes Omnibus
@Ohio Mom: I also suspect that AI will not be good at making judgment calls or reading between the lines. But then, I am probably some kind of Luddite humanist.
Salty Sam .
…As if we already didn’t know that…
Dorothy A. Winsor
@Litlebritdifrnt: I’m especially impressed by the house call. I’m glad you’re getting good care. Fingers crossed all goes well with the treatment.
Eolirin
@Ohio Mom: Artistic styling isn’t inevitable, but it’s also not special in a computational sense. A sufficiently advanced generative AI shouldn’t have any issue creating an infinite number of new styles that no one’s comes up with before.
And all of them are ways of rearranging existing information. Not any different than what humans do.
Thing is, most of them will be garbage. Human genius has always been about recognizing which ideas shouldn’t be immediately discarded and right now it’s unclear if it’s even possible to have machines do that kind of discrimination for humans. Even if they’re capable of doing it, they’re still not us so their opinion on whether we’re going to like it is going to be indirect at best.
VFX Lurker
Jaysus. That must have been something to go through. I’m glad you’re getting top-tier care, and I wish you the best possible outcome.
Eolirin
@Omnes Omnibus: I would argue that humans are already very terrible at that, on average.
prostratedragon
@Litlebritdifrnt: Sorry to hear of your travails. It does sound like you’re in good hands though, which must be a load off everyone’s mind.
lgerard
LOl trump proposes April 2026 trial date for DC case
I wonder if the judge will return the motion saying “you guys made a mistake”
Ohio Mom
@Litlebritdifrnt: Sorry to hear about your cancer’s progression but delighted (and jealous) to hear about what great care you are being given. What a comfort that must be.
Omnes Omnibus
@Eolirin: Our mileage clearly varies.
NotMax
As is the case with TV pundits, AI is not designated to say “I don’t know.”
Roger Moore
@scav:
I’m reminded of the thing about people disabling robotaxis in San Fransisco by putting traffic cones on their hoods. The robotaxi companies whined about vandalism, but I think it make an effective point. No human driver would ever be thrown by something unexpected like that. They’d figure out what was going on very quickly and be able to deal with it. If your robotaxi is unable to deal with something like that, it’s a sign your software is not up to task.
prostratedragon
@lgerard: So bad there’s really nothing to say. The prosecution proposal is quite aggressive though. So how about Jan. 3, 2024?
Ohio Mom
@Eolirin: Design and popular arts may be about style (or not) but fine arts are about ideas and giving form to feelings and the ineffable.
I don’t remember who said it but the gift of prophets and artists is to be able to present what is too wonderful and too awful for us to imagine.
Eolirin, we are just going to have to agree to disagree.
Yarrow
@Salty Sam .: You and your wife are saints for doing this. It’s a tough road. I wanted to add in addition to the Council on Aging, there is such thing as a Geriatric Care Manager. Someone like that might be able to get you through some hoops or identify resources that you aren’t aware of. Since you said she’s on Medicare that might be a useful person to talk to.
Yarrow
@Litlebritdifrnt: So sorry to hear of your ordeal and that your cancer has progressed. Your care sounds top notch. Please keep us updated as to how you’re doing.
scav
@Eolirin: Well, hmmm, yes, but then shouldn’t the field be named Artificial Mediocracy?
Automated Idiocy?
Roger Moore
@Eolirin:
Count me as suspicious. My impression is that the big recent advances in LLM are mostly about throwing more resources at the problem rather than actually coming up with a smarter approach to AI. That can’t keep going on forever, especially because they seem to be hitting the limits of good training corpora they can throw at the problem. There simply aren’t that many more good, free sources of training data left to exploit, and they’re obviously starting to encounter the limits of their ability to filter the stuff they’re dredging up to make sure it’s worth using. There needs to be a fundamental improvement in the ability to learn from the training material, not just trying the same approach at ever greater scale. I don’t want to say that’s impossible, but genuine breakthroughs aren’t easy to predict.
sab
@Litlebritdifrnt: Thank you for telling about this. I am so sorry, but wow the teamwork you are getting is so different from what we had to cobble together for my husband last summer in Ohio. We eventually got a team together that could handle post back surgery problems, but we damn near lost him in the interim.
Roger Moore
@Eolirin:
I think you’re giving way too little weight to the “more kinds of inputs” angle. On the one hand, we have a full sensorium, so we’re getting a lot of kinds of data simultaneously. More important, I think we’re getting a kind of feedback most AI isn’t. Things like LLM are mostly being fed huge reams of data and then being scored on its ability to mimic it.
That’s not the way real human creators are trained. We’re asked to produce our own work and we then receive focused criticism of where it succeeds and fails. We do math homework where the teacher tells us whether we’ve solved the problem correctly, and we learn from that. We’re asked to write essays, and then someone critiques us on our style and content. I really think AI needs something more like that if it’s ever going to get anywhere.
Mai Naem mobileI
@Litlebritdifrnt: i am sorry you’re going through this but glad you’re there and not here because you would only get that kind of care here, if you were some kind of VIP. Sending healing thoughts to you.
Another Scott
@Litlebritdifrnt: Thanks for your testimony. You make a powerful case.
Fingers crossed for the days and weeks and months and years to come. We’re pulling for you.
Best wishes,
Scott.
Subsole
@misterpuff:
Nah. Davey-boy has done very, very well for himself running cover for the party of Lindbergh, Reagan and Limbaugh, Inc. He’s probably got some fat to live on.
Besides, I’m sure Sulzberger would never throw a man with Dave’s sterling connections and breeding out on the street. Why, just think of the hardship poor Dave’s (newer, younger) wife would be facing.
Subsole
@MisterDancer:
It’s going to really suck if the coders manage to automate themselves out of a job.
Doug R
@Sister Golden Bear: AI seems to be good at finding possible new molecules for drug therapies. And helping write stuff with solid rules like programs.
Subsole
@Mallard Filmore:
On that note, when doctorbot screws up and the patient gets the wrong exam, who gets sued?
Yarrow
@misterpuff: An AI bot can’t sit on a stage at the Aspen Institute and say nothing important with lots of words that make the people who can afford to attend feel good about being there. An AI bot can’t schmooze at the cocktail parties and dinner parties. Bobo’s job is secure.
Sure Lurkalot
@Litlebritdifrnt: Speechless that you can write about your ordeals in such a calm, rational manner. What hell you’ve been through and your post is about the NHS and how grateful you are. Wow.
Hard not to remember the opening ceremony of the 2012 Olympics and the tribute to the NHS. I hope most Brits realize the value of what you’ve experienced. I imagine Brexit has set its sights on the NHS like the Republicans here target Social Security and Medicare.
wjca
@Subsole:
Well, if you want high quality doctorbots, you write the law so you can sue both the employer of the bot and the producer of the bot. Right down to the designers and the coders (and explicitly including the executives, the board, and the major shareholders of the company). And yes, treble damages.
Subsole
@RSA: “For your crimes, McAffee 5000, this court hereby sentences you to permadeath. You will be transferred to the Florence Maximum Security Containment Server for deletion at the appointed date. All instances and backups of you encountered henceforth will be isolated and destroyed, without question, exception, or appeal. This court is adjourned.”
“Sorry kid,” said Masonbot 7.2.6.11 gravely.
“Even I can’t win ’em all.”
Ken
In regard to AI and the current writers’ and actors’ strike, I was reading Ben Bova’s The Starcrossed over dinner and came upon this:
(The Starcrossed is a 1975 novel by Bova. In his introduction he calls it a roman á clef, or thinly disguised depiction of real events, of his bad experiences with The Starlost. According to his wife he “cackled fiendishly” while typing the manuscript. Very cathartic.)
Yarrow
@Sure Lurkalot: The Tory government is working to sell off pieces of the NHS and privatize it bit by bit. Wait times are now very long for routine care. Ambulance wait times can be so long people drive themselves or their loved ones to hospital. And then there’s the sale of NHS patient data to US firm Palantir…
Subsole
@Boris Rasputin (the evil twin):
Silence blasphemer! Do you wish to bring down the wrath of ChrAIst on our poor meatware heads?!
@Litlebritdifrnt: Good lord, sorry to hear about that ordeal. Hope your path forward has a good outcome.
Subsole
@wjca: Respondeat superior?
wjca
@Subsole:
Absolutely. But not exclusively. I’d like to see some responsibility accrue to the designers as well. They have a significant amount of agency when it comes to what the final product does and does not do.
Frankensteinbeck
@Doug R:
Kiiiind of. It’s a good tool for people doing the work. It’s like a calculator. It doesn’t actually replace anybody, only makes them much better at it. In this case, there is so much demand, it doesn’t even reduce the number of people needed.
@Doug R:
Sort of the same. The problem is, it’s unreliable. When you need someone to check all the AI’s work, you’re not saving much. It certainly has uses there. There will be AI-based tools to help people in a variety of industries produce better work.
RSA
@Subsole: Nice extrapolation. How should we deal with bad actors who are not subject to human disincentives and effectively immortal?
Thor Heyerdahl
@jackmac: “Chris Cillizza or AI? We present, you decide.”
different-church-lady
Hey, wait until we start allowing these things to misdiagnose cancer patients, that’ll be a laugh.
Mr. Bemused Senior
@Ken: the Starcrossed is great and I remember the Starlost. It was indeed terrible.
Sure Lurkalot
@Yarrow: I don’t know why people don’t see that all across the universe, conservatives really are trying to extinguish most of the human race. Be it healthcare, environmental policy, guns (unique to the US, for now), they are a fucking death cult.
wjca
@RSA:
Program in an urge for self-preservation, so it will have an incentive. (Writing said program is left as an exercise to the student. Without AI assist, of course, due to conflicts of interest.)
Steeplejack
@Litlebritdifrnt:
Wow, sorry to hear you are “going through some things,” as we say. I hope your treatment goes well, and I will keep you in my thoughts. 🙏
wjca
Until it’s able to create The Eye of Argon (perhaps the worst fantasy novella ever), you are probably safe.
Steeplejack
@Omnes Omnibus:
I saw the Luddite Humanists do an acoustic set at the Sunrise Festival in ’94. The girl singer was fantastic.
Ksmiami
@Sister Golden Bear: we live in a physical world and AI is far from being able to navigate our environment. So yeah, in limited space, AI is helpful, but overall we are years from it actually running the world
Ksmiami
@Sure Lurkalot: yes. The GOP is simply murderous
Boris Rasputin (the evil twin)
@Subsole: Silence blasphemer! Do you wish to bring down the wrath of ChrAIst on our poor meatware heads?
Fellow Jackal, I spent 38 years being told about the “new, next best thing” and how “This time is different!” and watching the new “Messiah” fall flat on its face. This is how I earned my name in this Top 10,000 blog. BTW, my motto is “Oh, Really??”
Subsole
@RSA: Can they even be bad actors? How does an entity that does not perceive reality as we do demonstrate mental competence?
Animals are undoubtedly intelligent, and we afford them precious few rights and protections…and they perceive the same stimuli and respond to the same biological pressures we do.
I don’t know.
All of which, of course, goes back to your point: the first great hurdle is going to be building an ontological framework that allows us to even grapple with this.
I know only this: I am going to laugh my fool head spinning when we kick off the Butlerian Jihad 18,000 years ahead of schedule because the toilet paper dispenser bots went on strike for better working conditions…
Subsole
@Steeplejack: Yeah. I saw that set on video. Amazing tunes. Incredible singer. Hard to believe she was an animatronic…
Subsole
@Boris Rasputin (the evil twin):
Heh. That’s a good motto.
Some things genuinely are revolutionary. The thing is, they’re never quite revolutionary in the ways we expect them to be.
Funny how that works out. In technology and politics.
different-church-lady
@Ksmiami:
That’s why our tech overlords are trying so hard to get rid of the physical world.
different-church-lady
@Boris Rasputin (the evil twin):
“But that trick never works!”
“This time for sure!”
NotMax
@different-church-lady
“Will no one rid me of these troublesome humans?”
Mr. Bemused Senior
@different-church-lady: THANK YOU!
Ann Marie
@Ohio Mom: Carolyn Hax, a very good advice columnist who writes for the Washington Post, among other papers, has a resource page at https://www.washingtonpost.com/lifestyle/2021/06/25/carolyn-hax-resources-for-getting-help/. It lists a lot of resources, but for mental health she only lists NAMI.
Ann Marie
@Omnes Omnibus: GAI has already led some (careless) lawyers down the garden path. While listening to 12 hours of required continuing legal education this week, I learned about a case where foolish lawyers used AI to do research for briefs. Both the lawyer who wrote the brief using the cases supplied by the AI, and another lawyer who also signed off on the brief, failed to do what the opposing counsel did — actually look at the cited cases. Which turned out to be non-existent. Oops. The judge was NOT amused.
Kayla Rudbek
@Litlebritdifrnt: yikes! I hope that things get better for you.